摘要
随着训练集规模的不断增大,支持向量机学习成为了密集型计算的过程。针对计算过程中存在占用内存大、寻优速度慢等问题,通过大量实验对分组训练和层叠训练两种并行SVM算法进行性能分析,给出层叠分组SVM并行算法,并利用MapReduce并行框架实现,解决了层叠训练模型效率低的问题。实验结果表明,采用这种学习策略,在保持精度损失较小的情况下,一定程度上减少了训练时间,提高了分类速度。
With the constant growing of training set scale,support vector machine learning becomes intensive computing process. In view of the problems in calculation process including large memory and slow optimisation,we analyse the performances of two parallel SVM algorithms of grouping training and cascade training through a great deal of experiments,and present the cascade-grouping SVM parallel algorithm,and implement it using MapReduce parallel framework,this solves the problem of low efficiency of cascade training model.Experimental results show that by using this learning strategy,the training time is reduced and the classification speed is improved both to a certain extent without big precision loss.
出处
《计算机应用与软件》
CSCD
2015年第3期172-176,共5页
Computer Applications and Software
基金
国家自然科学基金项目(61363052)
内蒙古自然科学基金项目(2010MS0913)
内蒙古自治区气象信息中心合作项目